On Mon, 20 Jan 2014 11:17:17 +0200
"Michael S. Tsirkin" <> wrote:
> On Mon, Jan 20, 2014 at 02:28:00PM +1030, Rusty Russell wrote:
> > Cornelia Huck <> writes:
> > > Move the ccw payload description to tables, instead of using
> > > C structures with gcc extensions or textual descriptions.
> >
> > I think it's horrible.
> >
> > Almost everyone reading this will understand C. And they can cut and
> > paste the code from the spec into their implementation, which helps
> > everyone.
> >
> > We'll just have to live with the lack of padding, and let's try not to
> > add any more packed-requiring structures. If we deprecate
> > virtio-balloon, the rest will be contained in CCW.
> >
> > Cheers,
> > Rusty.
>
> For CCW, how about my idea of adding optional padding explicitly?
> In the compatibility section, we can explicitly specify the
> the length.
> Drivers can use offsetof(padding) and avoid packed attribute.
Adding compiler/language-dependend work-arounds (either the packed
attribute or the manual padding) to a generic standard specification
sounds really wrong to me ... but if Rusty really does not like the
tables, I also do not have a better idea here.
Thomas
